12. Statistical Modelling and Optimal Estimation Flashcards
1
Q
What are the basics to the MDL Philosophy
A
2
Q
Why do we need principles for model selection and inference?
A
Avoiding overfitting, balancing model complexity, interpretability
3
Q
What is MDL rooted in?
A
Rooted in Shannon’s information theory to balance model complexity with
data fit.
4
Q
Name some reasons why MDL is attractive
A
. Occam’s Razor
2. No overfitting
3. Bayesian interpretation
4. No need for ’underlying truth’
5. Predictive interpretation
5
Q
What is the Two-Part MDL principle
A
6
Q
A
